Specific rubric edits (examples, context, bias reduction) raise human-LLM rater agreement in essay scoring and instruction following while complexity and conservative aggregation lower it.
Also, papers receive a score based on the work the student did complete even if they seem to be unfinished
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Quantifying the Statistical Effect of Rubric Modifications on Human-Autorater Agreement
Specific rubric edits (examples, context, bias reduction) raise human-LLM rater agreement in essay scoring and instruction following while complexity and conservative aggregation lower it.